Machine learning frameworks, with some demo applications and analysis.
- mlp_framework_vanilla.py - Complete multilayer
perceptron neural network framework, written in vanilla Python for
educational purposes. It is well commented and wildly inefficient. 🙃
-
Demo: XOR and linear separability - This notebook discusses the XOR function and linear separability. We construct a multi-layer perceptron that learns XOR, and then analyze the trained network and graph its decision boundaries.
-
Demo: Digit classifier - This notebook trains a multi-class digit classifier using the MNIST dataset, and features a couple of interesting visualizations. Planned addition: an interactive ipynb widget for drawing digits with real-time visualization of the predicted probability distribution.
-